"Your hands already know."
An end-to-end somatic wellness system from hardware sensor to mobile app, designed to help users catch hidden stress before it becomes burnout.
About
In this thesis project, Orbit, I designed a somatic wellness system that bridges the gap between physical stress signals and emotional awareness for high-functioning, anxious professionals. Moving from secondary research and 700+ raw data points to a hi-fi Figma prototype and a functional hardware companion, I led the end-to-end process from problem framing to a pressure-sensitive physical device paired with a mobile app. My primary focus was on designing the "somatic feedback loop", translating raw grip-tension data into a warm, narrative-driven reflection experience, while leveraging AI tools throughout research synthesis, visual generation, and rapid prototyping to accelerate iteration.
Impact
The redesigned system successfully bridged the gap between cold, clinical data tracking and empathetic, judgment-free self-awareness. Two rounds of usability testing (SUS score of 83.2, in the excellent range) showed that the no-score, no-grade narrative approach was the most universally valued design decision, particularly among metric-anxious users who previously felt judged by traditional wellness trackers. By replacing data dashboards with "Traces" personal narrative letters, the design lowered the emotional barrier to self-reflection and helped users catch hidden stress before it reached a breaking point. This approach not only improved usability but also repositioned the role of wellness technology from surveillance to companionship, offering a new model for how AIoT products can support mental health.
Timeline
Aug 2025 – May 2026, 9 months
Task
Secondary & Primary Research
Design System
Hi-Fi Prototyping
Usability Testing (2 rounds)
AI-Assisted Rapid Prototyping
Team
Wenliang Huang
The Problem
We live in the age of the "Quantified Self", tracking steps, calories, and heart rate with clinical precision. Yet we've never felt more disconnected from what we're actually feeling. Global stress has hit a record high of 44% (Gallup, 2023), while our ability to sense what's happening inside our own bodies keeps declining.
The deeper issue: emotional states aren't simple numbers. They're complex, layered signals, part physical, part psychological. But most wellness tools flatten all of that into a single score, telling you that something changed, never what you're feeling or why.
Problem Statement: High-functioning people rely on 'Quantified Self' tools that track cold numbers, but these metrics often cause 'Metric Anxiety' rather than healing. The result is a dangerous "Mind-Body Gap" people believe they're fine, while their bodies are silently storing stress their minds haven't caught up to, until it surfaces as a headache, exhaustion, or burnout.
Research → AI Integration Point #1
700+ raw data points from surveys and interviews were distilled through Affinity Mapping into 108 Themes → 58 Key Insights → 4 Opportunity Areas → 3 HMWs — the "North Star" for the entire design phase.
I treated Claude and Google AI Studio as research partners, using them to cluster raw data, surface recurring patterns, and organize themes at a speed manual synthesis couldn't match, compressing weeks of analysis into days. Every AI-assisted cluster was then manually reviewed and validated against the original transcripts, ensuring the insights stayed grounded in what participants actually said.
This is what AI fluency looks like in practice for me: not offloading judgment to a tool, but using it to accelerate the path from raw data to actionable insight, so more time goes into interpreting what the data means, not organizing it.
Thesis / Solution Direction
Orbit replaces cold, clinical tracking with a tangible physical object, acting less like a monitoring tool and more like a warm friend. By pairing a pressure-sensitive hand-held ball with a supportive companion app, Orbit uses tactile interaction to help users catch hidden stress before it reaches a breaking point, moving beyond abstract metrics to align mental perception with actual bodily reality.
Three design principles guided every decision:
Bridging the Mind-Body Gap — Grip strength becomes a source of physiological truth, bypassing the mind's excuses and speaking directly to what the body already knows.
From Numbers to Narrative — Instead of a dashboard, Orbit offers "Traces": gentle, letter-style reflections that witness a user's week rather than score it.
Bio-Responsive Visuals — The interface itself "breathes," shifting in real time with the user's grip, so the screen and the body move in sync rather than in opposition.
The Solution
Orbit is a two-part system: a pressure-sensitive hand-held ball that reads grip tension in real time, paired with a companion app that turns that raw signal into something a person can actually reflect on. The ball detects emotional intensity through grip force, glows to mirror the user's current state, and syncs offline data automatically once reconnected while the app translates that physical data into check-ins, weekly summaries, and guided sessions.
Designing an AI-Assisted Reflection Layer
The core design challenge wasn't collecting data, it was deciding what to do with it. Raw grip-tension readings mean nothing to a stressed user staring at a chart. So instead of a dashboard, I designed "Traces": a feature that reframes the user's week as a personal, letter-style narrative rather than a set of scores.
This is where I designed for AI-assisted decision-making, not just AI-generated content. The system takes two inputs what the user says they feel, and what their grip actually revealed and synthesizes them into a short, empathetic reflection: acknowledging effort, surfacing a pattern ("your Monday mornings run tighter than your Fridays"), and gently pointing toward one small next step. The goal was to help users move from data to a decision not "here's your score," but "here's what this means, and here's what might help" directly translating a physiological signal into something actionable, without requiring the user to interpret it themselves.
Prototyping With AI, Not Just Designing For It
Beyond the interface, I used Claude Code and Antigravity to prototype the system's actual functionality generating the React/JavaScript for the interactive prototype and implementing Web Bluetooth logic to pull real-time sensor data from the physical device into the app. This meant I wasn't just mocking up what "grip-to-emotion-glow" might look like in Figma, I could hold the ball, squeeze it, and watch the interface respond live, using AI to compress the distance between a design idea and a working, testable prototype.
Every AI-generated output from code to visual assets was reviewed, debugged, and integrated by hand; AI accelerated iteration speed, not design judgment.
Design System
Usability Testing & Iteration
Every design decision in Orbit's second iteration was validated through structured usability testing, not assumption.
Two rounds of moderated usability testing were conducted with 14 participants across three user segments (Hidden Tension, Metric-Aware, Somatic-Curious). Round 1 produced a SUS score of 83.2 in the "excellent" range, validating the core direction: the conversational tone, the no-score philosophy, and the emotional check-in flow all landed well. But the same round surfaced 4 critical issues, each backed by direct participant quotes:
Facial recognition defaulted on without consent, causing 8/14 participants to feel surveilled and misclassified, someone who looks "composed at work" isn't necessarily calm.
The "Analyze" tab label felt clinical, breaking the app's warm tone and causing hesitation before entry.
"Your Path" lacked visual hierarchy, with 6/14 participants scrolling past it entirely
Only six emotion categories left 7/14 participants unable to name what they were actually feeling
Each issue was addressed directly, facial recognition was switched to a manual-first, opt-in flow; "Analyze" was renamed to something warmer; "Your Path" was given a stronger visual entry point; and a secondary intensity-and-context screen was added to capture emotional nuance the six categories missed. This before/after loop, not intuition, shaped every subsequent design decision.
Version 1
Version 2
Cross-Functional Collaboration
While Orbit was an independent thesis project, it wasn't built in isolation. I worked closely with my thesis committee throughout development, treating their expertise the way I would treat cross-functional partners on a product team.
Professor Clark served as my hardware advisor, guiding decisions on grip-sensor selection, Bluetooth data pipeline architecture, and the physical feasibility of translating raw pressure readings into meaningful emotional signals, pushing me to validate that the technology could actually support the experience I was designing.
Professor Jeannie guided my research and usability testing methodology, helping me structure both rounds of testing, define the right metrics (like SUS scoring), and interpret participant feedback rigorously rather than reading into results I wanted to see.
Regular check-ins with both advisors meant sharing work early and often, defending design decisions with evidence, and iterating based on direct feedback, the same rhythm of "share early, iterate often" that defines how I'd want to work with PM and engineering partners in a team setting.
Impact & Reflection
Impact
Orbit's redesigned system successfully closed the gap between cold, automated data tracking and genuine self-awareness. Across two rounds of usability testing (SUS 83.2, in the excellent range), the no-score, no-grade narrative approach emerged as the most universally valued design decision, particularly for metric-anxious users who previously felt judged by traditional wellness trackers. By replacing dashboards with "Traces," personal narrative reflections, the design lowered the emotional barrier to self-reflection and helped users catch hidden stress before it escalated into burnout.
Beyond the individual product, Orbit demonstrates a broader shift in how I think about designing with AI: not as a shortcut to output, but as a way to compress the distance between insight and action, whether that's synthesizing 700+ research data points into actionable themes, or turning a raw sensor reading into a decision a user can actually act on. That same principle, translating complexity into something clear enough to build trust and prompt action is what I look for in every product problem, regardless of the domain.
Reflection
This project pushed me to grow in three ways I'd bring directly into a product design role. First, AI fluency became a design skill, not just a technical one, knowing when to let AI accelerate synthesis and prototyping, and when a decision needed human judgment, sharpened how I think about responsible, effective AI-assisted design. Second, working across hardware constraints, interaction design, and two rounds of rigorous usability testing taught me to hold both craft and evidence accountable to each other, a beautiful interface means little if testing shows it doesn't build trust. Third, regular critique from my committee reinforced a habit I want to carry forward: share work early, defend decisions with data, and treat feedback as fuel for iteration, not a detour from it.
Ultimately, Orbit reframed my understanding of what good design does: it doesn't just make information legible, it makes people feel witnessed, not evaluated. That's the standard I want to bring to every product I help build.